大版本更新。

.gitignore

@@ -2,5 +2,6 @@ cmake-build-debug
 build
 .idea
 Mark
+armor_box_photo
 tools/TrainCNN/.idea
 tools/TrainCNN/__pycache__

armor/include/armor_finder/armor_finder.h

@@ -26,6 +26,7 @@ private:
     const uint8_t &enemy_color;
     State state;
     cv::Rect2d armor_box;
+    int boxid;
     cv::Ptr<cv::Tracker> tracker;
     cv::Mat src_gray;
 

armor/include/show_images/show_images.h

@@ -9,11 +9,13 @@
 #include <opencv2/highgui.hpp>
 #include <opencv2/imgproc.hpp>
 #include <armor_finder/armor_finder.h>
+#include <map>
 
+extern std::map<int, string> id2name;
 void showArmorBoxVector(std::string windows_name, const cv::Mat &src, const std::vector<cv::Rect2d> &armor_box);
-void showArmorBox(std::string windows_name, const cv::Mat &src, cv::Rect2d armor_box);
+void showArmorBox(std::string windows_name, const cv::Mat &src, cv::Rect2d armor_box, int boxid);
 void showContours(std::string windows_name, const cv::Mat &src, const std::vector<LightBlob> &light_blobs);
 void showArmorBoxClass(std::string window_names, const cv::Mat &src, vector<cv::Rect2d> boxes[10]);
-
+void showCuoWeiDu(const cv::Mat &src, const std::vector<LightBlob> &light_blobs);
 
 #endif /* _SHOW_IMAGES_H_ */

armor/src/armor_finder/armor_finder.cpp

@@ -14,7 +14,8 @@ ArmorFinder::ArmorFinder(uint8_t &color, Serial &u, string paras_folder, const u
             state(STANDBY_STATE),
             classifier(std::move(paras_folder)),
             contour_area(0),
-            use_classifier(use)
+            use_classifier(use),
+            boxid(-1)
             {
 }
 
@@ -25,11 +26,11 @@ void ArmorFinder::run(cv::Mat &src) {
     cv::cvtColor(src_use, src_gray, CV_RGB2GRAY);
 
     if(show_armor_box){
-        showArmorBox("box", src, armor_box);
+        showArmorBox("box", src, armor_box, boxid);
         cv::waitKey(1);
     }
-//    stateSearchingTarget(src_use);
+    stateSearchingTarget(src_use);
-//    return;
+    return;
     switch (state){
         case SEARCHING_STATE:
             if(stateSearchingTarget(src_use)){

armor/src/armor_finder/classifier/classifier.cpp

@@ -300,6 +300,7 @@ int Classifier::operator()(const cv::Mat &image) {
     vector<MatrixXd> sub = {b, g, r};
     vector<vector<MatrixXd>> in = {sub};
     MatrixXd result = calculate(in);
+//    cout << result << "==============" <<endl;
     MatrixXd::Index minRow, minCol;
     result.maxCoeff(&minRow, &minCol);
     return minRow;

armor/src/armor_finder/state_machine/searching_state/searching_state.cpp

@@ -168,13 +168,42 @@ bool lengthRatioJudge(const LightBlob &light_blob_i, const LightBlob &light_blob
             && light_blob_i.length / light_blob_j.length > 0.5);
 }
 
+/* 判断两个灯条的错位度，不知道英文是什么！！！ */
+bool CuoWeiDuJudge(const LightBlob &light_blob_i, const LightBlob &light_blob_j){
+    float angle_i = light_blob_i.rect.size.width > light_blob_i.rect.size.height ? light_blob_i.rect.angle:
+                    light_blob_i.rect.angle - 90;
+    float angle_j = light_blob_j.rect.size.width > light_blob_j.rect.size.height ? light_blob_j.rect.angle:
+                    light_blob_j.rect.angle - 90;
+    float angle = (angle_i+angle_j)/2.0/180.0*3.14159265459;
+    if(abs(angle_i-angle_j)>90){
+        angle += 3.14159265459/2;
+    }
+    Vector2f orientation(cos(angle), sin(angle));
+    Vector2f p2p(light_blob_j.rect.center.x-light_blob_i.rect.center.x, light_blob_j.rect.center.y-light_blob_i.rect.center.y);
+    return abs(orientation.dot(p2p)) < 20;
+}
+
+bool boxAngleJudge(const LightBlob &light_blob_i, const LightBlob &light_blob_j){
+    float angle_i = light_blob_i.rect.size.width > light_blob_i.rect.size.height ? light_blob_i.rect.angle:
+                    light_blob_i.rect.angle - 90;
+    float angle_j = light_blob_j.rect.size.width > light_blob_j.rect.size.height ? light_blob_j.rect.angle:
+                    light_blob_j.rect.angle - 90;
+    float angle = (angle_i+angle_j)/2.0;
+    if(abs(angle_i-angle_j)>90){
+        angle += 90.0;
+    }
+    return (-120.0<angle && angle<-60.0) || (60.0<angle && angle<120.0);
+}
+
 bool isCoupleLight(const LightBlob &light_blob_i, const LightBlob &light_blob_j, uint8_t enemy_color) {
     return light_blob_i.BlobColor == enemy_color &&
            light_blob_j.BlobColor == enemy_color &&
            lengthRatioJudge(light_blob_i, light_blob_j) &&
            lengthJudge(light_blob_i, light_blob_j) &&
-           heightJudge(light_blob_i, light_blob_j) &&
+//           heightJudge(light_blob_i, light_blob_j) &&
-           angelJudge(light_blob_i, light_blob_j);
+           angelJudge(light_blob_i, light_blob_j) &&
+           boxAngleJudge(light_blob_i, light_blob_j) &&
+           CuoWeiDuJudge(light_blob_i, light_blob_j);
 
 }
 
@@ -195,8 +224,8 @@ static bool findArmorBoxes(LightBlobs &light_blobs, std::vector<cv::Rect2d> &arm
             double min_x, min_y, max_x, max_y;
             min_x = fmin(rect_left.x, rect_right.x) - 4;
             max_x = fmax(rect_left.x + rect_left.width, rect_right.x + rect_right.width) + 4;
-            min_y = fmin(rect_left.y, rect_right.y) - 0.3*(rect_left.height+rect_right.height)/2.0;
+            min_y = fmin(rect_left.y, rect_right.y) - 0.5*(rect_left.height+rect_right.height)/2.0;
-            max_y = fmax(rect_left.y + rect_left.height, rect_right.y + rect_right.height) + 0.3*(rect_left.height+rect_right.height)/2.0;
+            max_y = fmax(rect_left.y + rect_left.height, rect_right.y + rect_right.height) + 0.5*(rect_left.height+rect_right.height)/2.0;
             if (min_x < 0 || max_x > 640 || min_y < 0 || max_y > 480) {
                 continue;
             }
@@ -251,13 +280,13 @@ void get_blob_color(const cv::Mat &src, std::vector<LightBlob> &blobs) {
     }
 }
 
-int prior_red[] = {0, 2, 3, 4, 1, 5, 7, 8, 9, 6};
+int prior_blue[] = {6, 0, 2, 3, 4, 5, 1, 13, 7, 9, 10, 11, 12, 8};
-int prior_blue[]= {5, 7, 8, 9, 6, 0, 2, 3, 4, 1};
+int prior_red[]= {13, 7, 9, 10, 11, 12, 8, 6, 0, 2, 3, 4, 5, 1};
 
 bool ArmorFinder::stateSearchingTarget(cv::Mat &src) {
     cv::Mat split, src_bin/*, edge*/;
     LightBlobs light_blobs, light_blobs_, light_blobs_real;
-    std::vector<cv::Rect2d> armor_boxes, boxes_number[10];
+    std::vector<cv::Rect2d> armor_boxes, boxes_number[14];
     armor_box = cv::Rect2d(0,0,0,0);
 
     cv::cvtColor(src, src_gray, CV_BGR2GRAY);
@@ -295,7 +324,8 @@ bool ArmorFinder::stateSearchingTarget(cv::Mat &src) {
     light_blobs_real = light_blobs;
     get_blob_color(src, light_blobs_real);
     if(show_light_blobs){
-        showContours("blobs_real", src, light_blobs_real);
+        showContours("light_blobs", src, light_blobs_real);
+//        showCuoWeiDu(src, light_blobs_real);
         cv::waitKey(1);
     }
 
@@ -319,6 +349,7 @@ bool ArmorFinder::stateSearchingTarget(cv::Mat &src) {
             for(auto id : prior_blue){
                 if(!boxes_number[id].empty()){
                     armor_box = boxes_number[id][0];
+                    boxid = id;
                     break;
                 }
             }
@@ -326,6 +357,7 @@ bool ArmorFinder::stateSearchingTarget(cv::Mat &src) {
             for(auto id : prior_red){
                 if(!boxes_number[id].empty()){
                     armor_box = boxes_number[id][0];
+                    boxid = id;
                     break;
                 }
             }
@@ -337,9 +369,17 @@ bool ArmorFinder::stateSearchingTarget(cv::Mat &src) {
         }
         if(show_armor_boxes){
             showArmorBoxClass("class", src, boxes_number);
+            for(int i=0; i<sizeof(boxes_number)/ sizeof(boxes_number[0]); i++){
+                for(auto &box : boxes_number[i]){
+                    char filename[100];
+                    sprintf(filename, PROJECT_DIR"/armor_box_photo/%s_%d.jpg", id2name[i].data(), time(nullptr)+clock());
+                    cv::imwrite(filename, src(box));
+                }
+            }
         }
     }else{
         armor_box = armor_boxes[0];
+        boxid = -1;
     }
     if(split.size() == cv::Size(320, 240)){
         armor_box.x *= 2;

armor/src/show_images/show_images.cpp

@@ -1,7 +1,26 @@
 #include <show_images/show_images.h>
+#include <log.h>
 
 using namespace cv;
 
+std::map<int, string> id2name = {
+        {-1, "NO"},
+        { 0, "B1"},
+        { 1, "B2"},
+        { 2, "B3"},
+        { 3, "B4"},
+        { 4, "B5"},
+        { 5, "B7"},
+        { 6, "B8"},
+        { 7, "R1"},
+        { 8, "R2"},
+        { 9, "R3"},
+        {10, "R4"},
+        {11, "R5"},
+        {12, "R7"},
+        {13, "R8"},
+};
+
 void showArmorBoxVector(std::string windows_name, const cv::Mat &src, const std::vector<cv::Rect2d> &armor_box) {
     static Mat image2show;
     if (src.type() == CV_8UC1) // 黑白图像
@@ -27,20 +46,25 @@ void showArmorBoxClass(std::string window_names, const cv::Mat &src, vector<cv::
     {
         image2show = src.clone();
     }
-    for(int i=0; i<10; i++){
+    for(int i=0; i<14; i++){
         if(!boxes[i].empty()){
             for(auto box : boxes[i]){
-                char buff[2] = {0};
-                buff[0] = i + '0';
                 rectangle(image2show, box, Scalar(0, 255, 0), 1);
-                putText(image2show, buff, Point(box.x+2, box.y+2), cv::FONT_HERSHEY_TRIPLEX, 1, Scalar(255,0,0));
+                if(i == -1)
+                    putText(image2show, id2name[i], Point(box.x+2, box.y+2), cv::FONT_HERSHEY_TRIPLEX, 1, Scalar(0,255,0));
+                else if(0<=i && i<7)
+                    putText(image2show, id2name[i], Point(box.x+2, box.y+2), cv::FONT_HERSHEY_TRIPLEX, 1, Scalar(255,0,0));
+                else if(7<=i && i<14)
+                    putText(image2show, id2name[i], Point(box.x+2, box.y+2), cv::FONT_HERSHEY_TRIPLEX, 1, Scalar(0,0,255));
+                else
+                    LOGE_INFO("Invalid box id:%d!", i);
             }
         }
     }
     imshow(window_names, image2show);
 }
 
-void showArmorBox(std::string windows_name, const cv::Mat &src, cv::Rect2d armor_box) {
+void showArmorBox(std::string windows_name, const cv::Mat &src, cv::Rect2d armor_box, int boxid) {
     static Mat image2show;
     if (src.type() == CV_8UC1) // 黑白图像
     {
@@ -50,6 +74,14 @@ void showArmorBox(std::string windows_name, const cv::Mat &src, cv::Rect2d armor
         image2show = src.clone();
     }
     rectangle(image2show, armor_box, Scalar(0, 255, 0), 1);
+    if(boxid == -1)
+        putText(image2show, id2name[boxid], Point(armor_box.x+2, armor_box.y+2), cv::FONT_HERSHEY_TRIPLEX, 1, Scalar(0,255,0));
+    else if(0<=boxid && boxid<7)
+        putText(image2show, id2name[boxid], Point(armor_box.x+2, armor_box.y+2), cv::FONT_HERSHEY_TRIPLEX, 1, Scalar(255,0,0));
+    else if(7<=boxid && boxid<14)
+        putText(image2show, id2name[boxid], Point(armor_box.x+2, armor_box.y+2), cv::FONT_HERSHEY_TRIPLEX, 1, Scalar(0,0,255));
+    else
+        LOGE_INFO("Invalid box id:%d!", boxid);
     imshow(windows_name, image2show);
 }
 
@@ -67,13 +99,49 @@ void showContours(std::string windows_name, const cv::Mat &src, const std::vecto
 
     for(const auto &light_blob:light_blobs)
     {
+        Scalar color;
         if(light_blob.BlobColor == BLOB_RED)
-            rectangle(image2show, light_blob.rect.boundingRect(), Scalar(0,0,255), 3);
+            color = Scalar(0,0,255);
-        if(light_blob.BlobColor == BLOB_BLUE)
+        else if(light_blob.BlobColor == BLOB_BLUE)
-            rectangle(image2show, light_blob.rect.boundingRect(), Scalar(255,0,0), 3);
+            color = Scalar(255,0,0);
         else
-            rectangle(image2show, light_blob.rect.boundingRect(), Scalar(0,255,0), 3);
+            color = Scalar(0,255,0);
+        cv::Point2f vertices[4];
+        light_blob.rect.points(vertices);
+        for (int j = 0; j < 4; j++){
+            cv::line(image2show, vertices[j], vertices[(j + 1) % 4], color, 2);
+        }
     }
     imshow(windows_name, image2show);
 }
 
+void drawCuoWeiDu(cv::Mat &src, const LightBlob &light_blob_i, const LightBlob &light_blob_j){
+    float angle_i = light_blob_i.rect.size.width > light_blob_i.rect.size.height ? light_blob_i.rect.angle:
+                    light_blob_i.rect.angle - 90;
+    float angle_j = light_blob_j.rect.size.width > light_blob_j.rect.size.height ? light_blob_j.rect.angle:
+                    light_blob_j.rect.angle - 90;
+    float angle = (angle_i+angle_j)/2.0/180.0*3.14159265459;
+    if(abs(angle_i-angle_j)>90){
+        angle += 3.14159265459/2;
+    }
+    Point2f orientation(cos(angle), sin(angle));
+    Vector2f p2p(light_blob_j.rect.center.x-light_blob_i.rect.center.x, light_blob_i.rect.center.y-light_blob_j.rect.center.y);
+    cv::line(
+            src,
+            (light_blob_j.rect.center+light_blob_i.rect.center)/2.0,
+            (light_blob_j.rect.center+light_blob_i.rect.center)/2.0 + 100*orientation,
+            Scalar(0,255,0),
+            2
+    );
+}
+
+void showCuoWeiDu(const cv::Mat &src, const std::vector<LightBlob> &light_blobs){
+    Mat image2show = src.clone();
+    for (int i = 0; i < light_blobs.size() - 1; ++i) {
+        for (int j = i + 1; j < light_blobs.size(); ++j) {
+            drawCuoWeiDu(image2show, light_blobs[i], light_blobs[j]);
+        }
+    }
+    imshow("CuoWeiDu", image2show);
+}
+

main.cpp

@@ -32,7 +32,7 @@ mcu_data mcuData = {
         ARMOR_STATE,
         0,
         1,
-        ENEMY_RED,
+        ENEMY_BLUE,
 };
 
 int main(int argc, char *argv[]) {
@@ -57,14 +57,9 @@ int main(int argc, char *argv[]) {
         WrapperHead *video_armor=nullptr;
         WrapperHead *video_energy=nullptr;
         if (from_camera) {
-            video_armor = new CameraWrapper(0, "armor");
+            video_armor = new CameraWrapper(0/*, "armor"*/);
             video_energy = new CameraWrapper(1, "energy");
         } else {
-//            string armor_video, energy_video;
-//            lastVideo(armor_video, PROJECT_DIR"/armor_video/");
-//            video_armor = new VideoWrapper(armor_video);
-//            lastVideo(energy_video, PROJECT_DIR"/energy_video/");
-//            video_energy = new VideoWrapper(energy_video);
             video_armor = new VideoWrapper("/home/sjturm/Desktop/valid_video/armor/65.avi");
             video_energy = new VideoWrapper("/home/sjturm/Desktop/valid_video/energy/121.avi");
         }

others/src/camera/camera_wrapper.cpp

@@ -76,9 +76,9 @@ bool CameraWrapper::init() {
 #elif defined(Linux)
     CameraSetAeState(h_camera, false);
     CameraSetExposureTime(h_camera, 10*1000);
-    CameraSetAnalogGain(h_camera, 20);
+    CameraSetAnalogGain(h_camera, 40);
     if(mode == 0){
-        CameraSetGain(h_camera, 100, 130, 112);
+        CameraSetGain(h_camera, 100, 100, 100);
         CameraSetLutMode(h_camera, LUTMODE_PRESET);
     }
 #endif

tools/TrainCNN/backward.py

@@ -5,8 +5,8 @@ from tqdm import tqdm
 import generate
 import forward
 import cv2
-import sys
 import numpy as np
+import mvsdk
 print("Finish!")
 
 def save_kernal(fp, val):
@@ -54,7 +54,7 @@ def save_para(folder, paras):
         save_bias(fp, paras[7])
 
 
-STEPS = 100000
+STEPS = 5000
 BATCH = 30
 LEARNING_RATE_BASE  = 0.01
 LEARNING_RATE_DECAY = 0.99
@@ -62,12 +62,16 @@ MOVING_AVERAGE_DECAY = 0.99
 
 
 def train(dataset, show_bar=False):
+    test_images, test_labels = dataset.all_test_sets()
+
     x = tf.placeholder(tf.float32, [None, generate.SRC_ROWS, generate.SRC_COLS, generate.SRC_CHANNELS])
     y_= tf.placeholder(tf.float32, [None, forward.OUTPUT_NODES])
-    nodes, vars = forward.forward(x, 0.001)
+    keep_rate = tf.placeholder(tf.float32)
+    nodes, vars = forward.forward(x, 0.01)
     y = nodes[-1]
 
-    ce  = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=y, labels=tf.argmax(y_, 1))
+#    ce  = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=y, labels=tf.argmax(y_, 1))
+    ce  = tf.nn.weighted_cross_entropy_with_logits(logits=y, labels=tf.argmax(y_, 1), pos_weight=1)
     cem = tf.reduce_mean(ce)
     loss= cem + tf.add_n(tf.get_collection("losses"))
 
@@ -87,72 +91,118 @@ def train(dataset, show_bar=False):
 
     correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
     accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
+
     config = tf.ConfigProto(gpu_options=tf.GPUOptions(allow_growth=True))
     with tf.Session(config=config) as sess:
         init_op = tf.global_variables_initializer()
         sess.run(init_op)
 
-        bar = tqdm(range(STEPS), dynamic_ncols=True)
+        bar = tqdm(range(STEPS), ascii=True, dynamic_ncols=True)
         for i in bar:
             images_samples, labels_samples = dataset.sample_train_sets(BATCH)
 
             _, loss_value, step = sess.run(
                 [train_op, loss, global_step],
-                feed_dict={x: images_samples, y_: labels_samples}
+                feed_dict={x: images_samples, y_: labels_samples, keep_rate:0.7}
             )
 
             if i % 100 == 0:
                 if i % 1000 == 0:
-                    test_samples, test_labels = dataset.sample_test_sets(1000)
+                    acc = sess.run(accuracy, feed_dict={x: test_images, y_: test_labels, keep_rate:1.0})
-                    acc = sess.run(accuracy, feed_dict={x: test_samples, y_: test_labels})
                 bar.set_postfix({"loss": loss_value, "acc": acc})
 
-        # video = cv2.VideoCapture("/home/xinyang/Desktop/Video.mp4")
-        # _ = True
-        # while _:
-        #     _, frame = video.read()
-        #     cv2.imshow("Video", frame)
-        #     k = cv2.waitKey(10)
-        #     if k == ord(" "):
-        #         bbox = cv2.selectROI("frame", frame, False)
-        #         print(bbox)
-        #         roi = frame[bbox[1]:bbox[1]+bbox[3], bbox[0]:bbox[0]+bbox[2]]
-        #         roi = cv2.resize(roi, (48, 36))
-        #         cv2.imshow("roi", roi)
-        #         cv2.waitKey(0)
-        #         roi = roi.astype(np.float32)
-        #         roi /= 255.0
-        #         roi = roi.reshape([1, 36, 48, 3])
-        #         res = sess.run(y, feed_dict={x: roi})
-        #         res = res.reshape([forward.OUTPUT_NODES])
-        #         print(np.argmax(res))
-        #     elif k==ord("q"):
-        #         break
-        # keep = True
-        # while keep:
-        #     n = input()
-        #     im = cv2.imread(n)
-        #     im = cv2.resize(im, (48, 36))
-        #     cv2.imshow("im", im)
-        #     if cv2.waitKey(0) == ord("q"):
-        #         keep = False
-        #     im = im.astype(np.float32)
-        #     im /= 255.0
-        #     im = im.reshape([1, 36, 48, 3])
-        #     res = sess.run(y, feed_dict={x: im})
-        #     res = res.reshape([forward.OUTPUT_NODES])
-        #     print(np.argmax(res))
 
-        test_samples, test_labels = dataset.sample_test_sets(100)
+
         vars_val = sess.run(vars)
-        save_para("/home/xinyang/Desktop/AutoAim/tools/para", vars_val)
+        save_para("/home/xinyang/Desktop/RM_auto-aim/tools/para", vars_val)
-        nodes_val = sess.run(nodes, feed_dict={x:test_samples})
+        print("save done!")
-        return vars_val, nodes_val, test_samples
+#        nodes_val = sess.run(nodes, feed_dict={x:test_images})
+#        return vars_val, nodes_val
+        DevList = mvsdk.CameraEnumerateDevice()
+        nDev = len(DevList)
+        if nDev < 1:
+            print("No camera was found!")
+            return
+
+        for i, DevInfo in enumerate(DevList):
+            print("{}: {} {}".format(i, DevInfo.GetFriendlyName(), DevInfo.GetPortType()))
+        i = 0 if nDev == 1 else int(input("Select camera: "))
+        DevInfo = DevList[i]
+        print(DevInfo)
+
+        # 打开相机
+        hCamera = 0
+        try:
+            hCamera = mvsdk.CameraInit(DevInfo, -1, -1)
+        except mvsdk.CameraException as e:
+            print("CameraInit Failed({}): {}".format(e.error_code, e.message) )
+            return
+
+        # 获取相机特性描述
+        cap = mvsdk.CameraGetCapability(hCamera)
+
+        # 判断是黑白相机还是彩色相机
+        monoCamera = (cap.sIspCapacity.bMonoSensor != 0)
+
+        # 黑白相机让ISP直接输出MONO数据，而不是扩展成R=G=B的24位灰度
+        if monoCamera:
+            mvsdk.CameraSetIspOutFormat(hCamera, mvsdk.CAMERA_MEDIA_TYPE_MONO8)
+        else:
+            mvsdk.CameraSetIspOutFormat(hCamera, mvsdk.CAMERA_MEDIA_TYPE_BGR8)
+
+        # 相机模式切换成连续采集
+        mvsdk.CameraSetTriggerMode(hCamera, 0)
+
+        # 手动曝光，曝光时间30ms
+        mvsdk.CameraSetAeState(hCamera, 0)
+        mvsdk.CameraSetExposureTime(hCamera, 30 * 1000)
+
+        # 让SDK内部取图线程开始工作
+        mvsdk.CameraPlay(hCamera)
+
+        # 计算RGB buffer所需的大小，这里直接按照相机的最大分辨率来分配
+        FrameBufferSize = cap.sResolutionRange.iWidthMax * cap.sResolutionRange.iHeightMax * (1 if monoCamera else 3)
+
+        # 分配RGB buffer，用来存放ISP输出的图像
+        # 备注：从相机传输到PC端的是RAW数据，在PC端通过软件ISP转为RGB数据（如果是黑白相机就不需要转换格式，但是ISP还有其它处理，所以也需要分配这个buffer）
+        pFrameBuffer = mvsdk.CameraAlignMalloc(FrameBufferSize, 16)
+
+        while (cv2.waitKey(1) & 0xFF) != ord('q'):
+            # 从相机取一帧图片
+            try:
+                pRawData, FrameHead = mvsdk.CameraGetImageBuffer(hCamera, 200)
+                mvsdk.CameraImageProcess(hCamera, pRawData, pFrameBuffer, FrameHead)
+                mvsdk.CameraReleaseImageBuffer(hCamera, pRawData)
+                # 此时图片已经存储在pFrameBuffer中，对于彩色相机pFrameBuffer=RGB数据，黑白相机pFrameBuffer=8位灰度数据
+                # 把pFrameBuffer转换成opencv的图像格式以进行后续算法处理
+                frame_data = (mvsdk.c_ubyte * FrameHead.uBytes).from_address(pFrameBuffer)
+                frame = np.frombuffer(frame_data, dtype=np.uint8)
+                frame = frame.reshape((FrameHead.iHeight, FrameHead.iWidth, 1 if FrameHead.uiMediaType == mvsdk.CAMERA_MEDIA_TYPE_MONO8 else 3) )
+
+                frame = cv2.resize(frame, (640,480), interpolation = cv2.INTER_LINEAR)
+                cv2.imshow("Press q to end", frame)
+                if (cv2.waitKey(1)&0xFF) == ord(' '):
+                    roi = cv2.selectROI("roi", frame)
+                    roi = frame[roi[1]:roi[1]+roi[3], roi[0]:roi[0]+roi[2]]
+                    print(roi)
+                    cv2.imshow("box", roi)
+                    image = cv2.resize(roi, (48, 36))
+                    image = image.astype(np.float32) / 255.0
+                    out = sess.run(y, feed_dict={x:[image]})
+                    print(out)
+                    print(np.argmax(out))
+                
+            except mvsdk.CameraException as e:
+                if e.error_code != mvsdk.CAMERA_STATUS_TIME_OUT:
+                    print("CameraGetImageBuffer failed({}): {}".format(e.error_code, e.message) )
+
+        # 关闭相机
+        mvsdk.CameraUnInit(hCamera)
+
+        # 释放帧缓存
+        mvsdk.CameraAlignFree(pFrameBuffer)
 
 
 if __name__ == "__main__":
-    print("Loading data sets...")
+    dataset = generate.DataSet("/home/xinyang/Desktop/box_cut")
-    dataset = generate.DataSet("/home/xinyang/Desktop/dataset/box")
-    print("Finish!")
     train(dataset, show_bar=True)
-    input("Press any key to end...")

tools/TrainCNN/cv_grab.py

@@ -0,0 +1,94 @@
+#coding=utf-8
+import cv2
+import numpy as np
+import mvsdk
+
+def main_loop():
+	# 枚举相机
+	DevList = mvsdk.CameraEnumerateDevice()
+	nDev = len(DevList)
+	if nDev < 1:
+		print("No camera was found!")
+		return
+
+	for i, DevInfo in enumerate(DevList):
+		print("{}: {} {}".format(i, DevInfo.GetFriendlyName(), DevInfo.GetPortType()))
+	i = 0 if nDev == 1 else int(input("Select camera: "))
+	DevInfo = DevList[i]
+	print(DevInfo)
+
+	# 打开相机
+	hCamera = 0
+	try:
+		hCamera = mvsdk.CameraInit(DevInfo, -1, -1)
+	except mvsdk.CameraException as e:
+		print("CameraInit Failed({}): {}".format(e.error_code, e.message) )
+		return
+
+	# 获取相机特性描述
+	cap = mvsdk.CameraGetCapability(hCamera)
+
+	# 判断是黑白相机还是彩色相机
+	monoCamera = (cap.sIspCapacity.bMonoSensor != 0)
+
+	# 黑白相机让ISP直接输出MONO数据，而不是扩展成R=G=B的24位灰度
+	if monoCamera:
+		mvsdk.CameraSetIspOutFormat(hCamera, mvsdk.CAMERA_MEDIA_TYPE_MONO8)
+	else:
+		mvsdk.CameraSetIspOutFormat(hCamera, mvsdk.CAMERA_MEDIA_TYPE_BGR8)
+
+	# 相机模式切换成连续采集
+	mvsdk.CameraSetTriggerMode(hCamera, 0)
+
+	# 手动曝光，曝光时间30ms
+	mvsdk.CameraSetAeState(hCamera, 0)
+	mvsdk.CameraSetExposureTime(hCamera, 30 * 1000)
+
+	# 让SDK内部取图线程开始工作
+	mvsdk.CameraPlay(hCamera)
+
+	# 计算RGB buffer所需的大小，这里直接按照相机的最大分辨率来分配
+	FrameBufferSize = cap.sResolutionRange.iWidthMax * cap.sResolutionRange.iHeightMax * (1 if monoCamera else 3)
+
+	# 分配RGB buffer，用来存放ISP输出的图像
+	# 备注：从相机传输到PC端的是RAW数据，在PC端通过软件ISP转为RGB数据（如果是黑白相机就不需要转换格式，但是ISP还有其它处理，所以也需要分配这个buffer）
+	pFrameBuffer = mvsdk.CameraAlignMalloc(FrameBufferSize, 16)
+
+	while (cv2.waitKey(1) & 0xFF) != ord('q'):
+		# 从相机取一帧图片
+		try:
+			pRawData, FrameHead = mvsdk.CameraGetImageBuffer(hCamera, 200)
+			mvsdk.CameraImageProcess(hCamera, pRawData, pFrameBuffer, FrameHead)
+			mvsdk.CameraReleaseImageBuffer(hCamera, pRawData)
+			
+			# 此时图片已经存储在pFrameBuffer中，对于彩色相机pFrameBuffer=RGB数据，黑白相机pFrameBuffer=8位灰度数据
+			# 把pFrameBuffer转换成opencv的图像格式以进行后续算法处理
+			frame_data = (mvsdk.c_ubyte * FrameHead.uBytes).from_address(pFrameBuffer)
+			frame = np.frombuffer(frame_data, dtype=np.uint8)
+			frame = frame.reshape((FrameHead.iHeight, FrameHead.iWidth, 1 if FrameHead.uiMediaType == mvsdk.CAMERA_MEDIA_TYPE_MONO8 else 3) )
+
+			frame = cv2.resize(frame, (640,480), interpolation = cv2.INTER_LINEAR)
+			cv2.imshow("Press q to end", frame)
+			roi = cv2.selectROI("roi", frame)
+			roi = frame[roi[1]:roi[1]+roi[3], roi[0]:roi[0]+roi[2]]
+			print(roi)
+			cv2.imshow("box", roi)
+
+			
+		except mvsdk.CameraException as e:
+			if e.error_code != mvsdk.CAMERA_STATUS_TIME_OUT:
+				print("CameraGetImageBuffer failed({}): {}".format(e.error_code, e.message) )
+
+	# 关闭相机
+	mvsdk.CameraUnInit(hCamera)
+
+	# 释放帧缓存
+	mvsdk.CameraAlignFree(pFrameBuffer)
+
+def main():
+	try:
+		main_loop()
+	finally:
+		cv2.destroyAllWindows()
+
+main()

tools/TrainCNN/forward.py

@@ -29,24 +29,25 @@ def max_pool_2x2(x):
 CONV1_KERNAL_SIZE = 5
 
 # 第一层卷积输出通道数
-CONV1_OUTPUT_CHANNELS = 8
+CONV1_OUTPUT_CHANNELS = 6
 
 # 第二层卷积核大小
 CONV2_KERNAL_SIZE = 3
 
 # 第二层卷积输出通道数
-CONV2_OUTPUT_CHANNELS = 16
+CONV2_OUTPUT_CHANNELS = 10
 
 # 第一层全连接宽度
 FC1_OUTPUT_NODES = 16
 
 # 第二层全连接宽度（输出标签类型数）
 FC2_OUTPUT_NODES = 15
+
 # 输出标签类型数
 OUTPUT_NODES = FC2_OUTPUT_NODES
 
 
-def forward(x, regularizer=None):
+def forward(x, regularizer=None, keep_rate=tf.constant(1.0)):
     vars = []
     nodes = []
 
@@ -71,16 +72,19 @@ def forward(x, regularizer=None):
     pool_shape = pool2.get_shape().as_list()
     node = pool_shape[1] * pool_shape[2] * pool_shape[3]
     reshaped = tf.reshape(pool2, [-1, node])
+    reshaped = tf.nn.dropout(reshaped, keep_rate)
 
-    fc1_w = tf.nn.dropout(get_weight([node, FC1_OUTPUT_NODES], regularizer), 0.1)
+    fc1_w = get_weight([node, FC1_OUTPUT_NODES], regularizer)
     fc1_b = get_bias([FC1_OUTPUT_NODES])
     fc1   = tf.nn.relu(tf.matmul(reshaped, fc1_w) + fc1_b)
+    fc1   = tf.nn.dropout(fc1, keep_rate)
     vars.extend([fc1_w, fc1_b])
     nodes.extend([fc1])
 
-    fc2_w = tf.nn.dropout(get_weight([FC1_OUTPUT_NODES, FC2_OUTPUT_NODES], regularizer), 0.1)
+    fc2_w = get_weight([FC1_OUTPUT_NODES, FC2_OUTPUT_NODES], regularizer)
     fc2_b = get_bias([FC2_OUTPUT_NODES])
-    fc2   = tf.nn.softmax(tf.matmul(fc1, fc2_w) + fc2_b)
+#    fc2   = tf.nn.softmax(tf.matmul(fc1, fc2_w) + fc2_b)
+    fc2   = tf.matmul(fc1, fc2_w) + fc2_b
     vars.extend([fc2_w, fc2_b])
     nodes.extend([fc2])
 

tools/TrainCNN/generate.py

@@ -2,10 +2,8 @@ import numpy as np
 import os
 import cv2
 import random
-from forward import OUTPUT_NODES
-import sys
-import os
 from tqdm import tqdm
+from forward import OUTPUT_NODES
 # 原图像行数
 SRC_ROWS = 36
 
@@ -24,7 +22,7 @@ class DataSet:
         self.test_labels = []
         self.generate_data_sets(folder)
 
-    def file2nparray(self, name, random=False):
+    def file2nparray(self, name):
         image = cv2.imread(name)
         image = cv2.resize(image, (SRC_COLS, SRC_ROWS))
         image = image.astype(np.float32)
@@ -42,16 +40,12 @@ class DataSet:
             files = os.listdir(dir)
             for file in tqdm(files, postfix={"loading id": i}, dynamic_ncols=True):
                 if file[-3:] == "jpg":
-                    try:
                     if random.random() > 0.2:
                         self.train_samples.append(self.file2nparray("%s/%s" % (dir, file)))
                         self.train_labels.append(self.id2label(i))
                     else:
                         self.test_samples.append(self.file2nparray("%s/%s" % (dir, file)))
                         self.test_labels.append(self.id2label(i))
-                    except:
-                        print("%s/%s" % (dir, file))
-                        continue
         self.train_samples = np.array(self.train_samples)
         self.train_labels = np.array(self.train_labels)
         self.test_samples = np.array(self.test_samples)
@@ -67,15 +61,6 @@ class DataSet:
             labels.append(self.train_labels[id])
         return np.array(samples), np.array(labels)
 
-    def sample_test_sets(self, length):
-        samples = []
-        labels = []
-        for i in range(length):
-            id = random.randint(0, len(self.test_samples)-1)
-            samples.append(self.test_samples[id])
-            labels.append(self.test_labels[id])
-        return np.array(samples), np.array(labels)
-
     def all_train_sets(self):
         return self.train_samples[:], self.train_labels[:]
 

tools/TrainCNN/grab.py

@@ -0,0 +1,111 @@
+#coding=utf-8
+import mvsdk
+
+def main():
+	# 枚举相机
+	DevList = mvsdk.CameraEnumerateDevice()
+	nDev = len(DevList)
+	if nDev < 1:
+		print("No camera was found!")
+		return
+		
+	for i, DevInfo in enumerate(DevList):
+		print("{}: {} {}".format(i, DevInfo.GetFriendlyName(), DevInfo.GetPortType()))
+	i = 0 if nDev == 1 else int(input("Select camera: "))
+	DevInfo = DevList[i]
+	print(DevInfo)
+
+	# 打开相机
+	hCamera = 0
+	try:
+		hCamera = mvsdk.CameraInit(DevInfo, -1, -1)
+	except mvsdk.CameraException as e:
+		print("CameraInit Failed({}): {}".format(e.error_code, e.message) )
+		return
+
+	# 获取相机特性描述
+	cap = mvsdk.CameraGetCapability(hCamera)
+	PrintCapbility(cap)
+
+	# 判断是黑白相机还是彩色相机
+	monoCamera = (cap.sIspCapacity.bMonoSensor != 0)
+
+	# 黑白相机让ISP直接输出MONO数据，而不是扩展成R=G=B的24位灰度
+	if monoCamera:
+		mvsdk.CameraSetIspOutFormat(hCamera, mvsdk.CAMERA_MEDIA_TYPE_MONO8)
+
+	# 相机模式切换成连续采集
+	mvsdk.CameraSetTriggerMode(hCamera, 0)
+
+	# 手动曝光，曝光时间30ms
+	mvsdk.CameraSetAeState(hCamera, 0)
+	mvsdk.CameraSetExposureTime(hCamera, 30 * 1000)
+
+	# 让SDK内部取图线程开始工作
+	mvsdk.CameraPlay(hCamera)
+
+	# 计算RGB buffer所需的大小，这里直接按照相机的最大分辨率来分配
+	FrameBufferSize = cap.sResolutionRange.iWidthMax * cap.sResolutionRange.iHeightMax * (1 if monoCamera else 3)
+
+	# 分配RGB buffer，用来存放ISP输出的图像
+	# 备注：从相机传输到PC端的是RAW数据，在PC端通过软件ISP转为RGB数据（如果是黑白相机就不需要转换格式，但是ISP还有其它处理，所以也需要分配这个buffer）
+	pFrameBuffer = mvsdk.CameraAlignMalloc(FrameBufferSize, 16)
+
+	# 从相机取一帧图片
+	try:
+		pRawData, FrameHead = mvsdk.CameraGetImageBuffer(hCamera, 2000)
+		mvsdk.CameraImageProcess(hCamera, pRawData, pFrameBuffer, FrameHead)
+		mvsdk.CameraReleaseImageBuffer(hCamera, pRawData)
+		
+		# 此时图片已经存储在pFrameBuffer中，对于彩色相机pFrameBuffer=RGB数据，黑白相机pFrameBuffer=8位灰度数据
+		# 该示例中我们只是把图片保存到硬盘文件中
+		status = mvsdk.CameraSaveImage(hCamera, "./grab.bmp", pFrameBuffer, FrameHead, mvsdk.FILE_BMP, 100)
+		if status == mvsdk.CAMERA_STATUS_SUCCESS:
+			print("Save image successfully. image_size = {}X{}".format(FrameHead.iWidth, FrameHead.iHeight) )
+		else:
+			print("Save image failed. err={}".format(status) )
+	except mvsdk.CameraException as e:
+		print("CameraGetImageBuffer failed({}): {}".format(e.error_code, e.message) )
+
+	# 关闭相机
+	mvsdk.CameraUnInit(hCamera)
+
+	# 释放帧缓存
+	mvsdk.CameraAlignFree(pFrameBuffer)
+
+def PrintCapbility(cap):
+	for i in range(cap.iTriggerDesc):
+		desc = cap.pTriggerDesc[i]
+		print("{}: {}".format(desc.iIndex, desc.GetDescription()) )
+	for i in range(cap.iImageSizeDesc):
+		desc = cap.pImageSizeDesc[i]
+		print("{}: {}".format(desc.iIndex, desc.GetDescription()) )
+	for i in range(cap.iClrTempDesc):
+		desc = cap.pClrTempDesc[i]
+		print("{}: {}".format(desc.iIndex, desc.GetDescription()) )
+	for i in range(cap.iMediaTypeDesc):
+		desc = cap.pMediaTypeDesc[i]
+		print("{}: {}".format(desc.iIndex, desc.GetDescription()) )
+	for i in range(cap.iFrameSpeedDesc):
+		desc = cap.pFrameSpeedDesc[i]
+		print("{}: {}".format(desc.iIndex, desc.GetDescription()) )
+	for i in range(cap.iPackLenDesc):
+		desc = cap.pPackLenDesc[i]
+		print("{}: {}".format(desc.iIndex, desc.GetDescription()) )
+	for i in range(cap.iPresetLut):
+		desc = cap.pPresetLutDesc[i]
+		print("{}: {}".format(desc.iIndex, desc.GetDescription()) )
+	for i in range(cap.iAeAlmSwDesc):
+		desc = cap.pAeAlmSwDesc[i]
+		print("{}: {}".format(desc.iIndex, desc.GetDescription()) )
+	for i in range(cap.iAeAlmHdDesc):
+		desc = cap.pAeAlmHdDesc[i]
+		print("{}: {}".format(desc.iIndex, desc.GetDescription()) )
+	for i in range(cap.iBayerDecAlmSwDesc):
+		desc = cap.pBayerDecAlmSwDesc[i]
+		print("{}: {}".format(desc.iIndex, desc.GetDescription()) )
+	for i in range(cap.iBayerDecAlmHdDesc):
+		desc = cap.pBayerDecAlmHdDesc[i]
+		print("{}: {}".format(desc.iIndex, desc.GetDescription()) )
+
+main()

tools/para/conv1_b

@@ -1,9 +1,7 @@
-8
+6
-0.026843265
+2.196893
-0.13687223
+0.07216131
-0.355584
+0.30069783
--2.171335
+-0.4587247
-2.0351274
+0.25476167
-1.8288306
+-0.07236218
--4.113487
--4.7374034

tools/para/conv2_b

@@ -1,17 +1,11 @@
-16
+10
-1.252942
+-0.20609309
-8.216776
+-0.22031759
--0.25801975
+0.1367356
-0.23331891
+0.3687642
--1.0068187
+0.41563538
--1.1067235
+0.56676525
--0.40771145
+-0.18027179
--0.43731463
+0.23183917
--1.4359887
+-0.42312288
--0.2637226
+-0.071102634
--0.41042513
-5.653234
-11.668375
--1.4110142
--0.1370871
--0.19704156

tools/para/fc1_b

@@ -1,17 +1,17 @@
 16
--0.20538531
+0.58729273
-7.367273
+-0.46309644
--0.18452525
+-0.16430101
-6.532006
+0.43460655
-25.536476
+-0.12165885
--0.18481636
+-0.23968913
--7.2863836
+2.1033192
--0.106642306
+-0.19900312
--13.070918
+-0.0075173783
--0.20218277
+0.05968375
-12.15478
+-0.13455966
--0.28686985
+-0.203078
--0.0753381
+3.4187536
--0.18774705
+-0.17911159
--0.45540679
+0.2670588
--0.81279093
+-0.58640796

tools/para/fc2_b

@@ -1,12 +1,16 @@
-11
+15
-3.028916
+3.3147552
-1.399315
+0.06590654
-12.311913
+-0.37396204
-1.9181013
+-0.22522521
--6.701019
+-0.9601034
-4.332221
+-0.9866448
--1.2238123
+-0.091494516
-3.367433
+0.08088531
--12.5565
+-0.87962383
--5.421737
+-0.5273214
-0.19371712
+-0.18194006
+-0.035499398
+-1.7873636
+0.48932117
+0.20472674

tools/para/fc2_w

@@ -1,178 +1,242 @@
 16
-11
+15
--6.6445126e-34
+0.032564595
--3.249375e-34
+0.2617493
-1.1159086e-33
+-0.32648245
-3.8638357e-34
+-0.09931708
-1.1999902e-33
+-0.24592394
--4.773489e-34
+-0.046472594
-6.573301e-35
+0.33926198
--2.7309886e-34
+-0.12294629
--3.5826868e-34
+0.35003394
--2.058676e-35
+-0.21551898
-2.9253375e-34
+-0.017696692
--0.028950384
+-0.14498983
-0.6513481
+-0.1035376
--0.090478554
+0.38845813
--0.20445836
+-0.025425002
--0.033577178
+0.009146354
-0.07797232
+-0.10837719
-0.54864347
+-0.26169685
--0.29817155
+0.24757256
--0.0972982
+0.12278806
--0.06013593
+0.173229
--0.12082546
+-0.13405079
-3.2078713e-35
+-0.12579814
-9.031606e-34
+-0.055770937
-2.9989992e-34
+-0.18405183
--1.8143521e-34
+0.28358203
--4.3935644e-34
+0.07445254
--7.0448736e-34
+0.23714246
-7.1789805e-35
+-0.13335316
--1.0773237e-33
+0.010074598
-4.1924878e-35
+8.8978056e-36
-2.6375152e-35
+-3.2673058e-35
-7.6904e-34
+-5.345727e-35
-0.7534392
+-1.2187582e-35
-0.041451767
+2.1464323e-35
--0.16353445
+6.242724e-35
--0.060047485
+-1.5923025e-35
-0.00937684
+8.5710344e-35
--0.046534266
+-1.0859142e-34
-0.12362613
+3.2036078e-35
--0.15848428
+9.889982e-35
--0.24788214
+9.5151974e-35
--0.10429883
+-2.9864197e-35
-0.07533859
+4.3109238e-35
--0.059416637
+-3.5075268e-35
--0.41185078
+0.056317
--0.17300163
+-0.041549474
--0.5048911
+-0.07867665
--0.27550554
+-0.18685594
--0.3164118
+0.0036230602
-0.41110015
+0.26440525
-0.5305193
+-0.040618088
-0.54406047
+-0.011269322
-0.41247433
+-0.037696317
-0.37498224
+0.01897098
--1.8530121e-34
+-0.12073718
-9.619048e-34
+0.017303245
--9.084177e-35
+0.33418366
--6.3287863e-34
+0.0023682562
--5.0406337e-34
+0.02849121
-6.430404e-34
+5.652079e-35
-3.175955e-34
+-2.3405746e-35
-4.679148e-34
+7.754459e-35
--9.705965e-35
+-7.9132345e-35
--1.4937167e-34
+-3.652418e-35
--3.7373778e-34
+2.7482412e-35
--1.8647024e-34
+-5.490258e-35
-2.157429e-34
+-4.0657551e-35
-9.178287e-35
+-8.77158e-35
-2.0542673e-34
+-1.6310674e-35
-4.1186567e-34
+6.9110407e-35
--1.4028581e-34
+2.8374646e-35
--1.9601842e-35
+7.249574e-35
-1.2199764e-34
+-8.326536e-36
-6.605314e-36
+-4.2208914e-35
-4.5839516e-35
+3.380163e-07
-4.1222883e-34
+0.00018250165
--5.649719e-34
+-4.091571e-13
-6.8534397e-34
+-5.742375e-20
-1.5284239e-34
+1.9097627e-29
-4.6217582e-35
+-1.0989738e-34
--2.6860813e-34
+4.988384e-06
-6.4033865e-34
+-6.4157313e-25
--1.9073337e-34
+0.0046886215
--4.5628154e-34
+-4.15727e-06
--2.5596114e-34
+8.565781e-06
-3.5286568e-34
+1.3159001e-08
--4.590898e-34
+-6.0661813e-27
--0.016765846
+0.003999361
-0.011994723
+4.6603424e-12
--0.26132298
+0.05875436
-0.52835166
+0.1978433
--0.21429977
+0.2357523
-0.047839653
+0.26039347
-0.0091085555
+0.29742035
--0.27048072
+0.23000301
-0.35106397
+0.22130986
--0.05962828
+0.32969925
--0.06534093
+-0.25387922
--5.9855516e-34
+-0.21843708
--3.8872762e-34
+-0.35505614
-1.4836724e-34
+-0.18760061
--3.7528057e-34
+-0.26636004
-9.244409e-35
+-0.3437664
-3.8288393e-34
+-0.31676972
-1.7450431e-34
+-2.9646424e-21
--2.1571653e-34
+1.6212045e-32
--8.635735e-34
+-4.972171e-35
--1.1816434e-33
+3.321333e-35
-2.75913e-34
+-3.0660306e-36
--0.11307323
+-1.2462985e-35
--0.05993526
+-6.15827e-35
--0.13786606
+-7.708171e-35
-0.0066387164
+-8.6527984e-35
-0.0024843283
+-9.63909e-35
-0.59352225
+-4.329017e-36
--0.13324556
+-1.6798441e-35
--0.275834
+6.4576066e-36
--0.13921
+1.0103299e-34
--0.023196468
+5.888647e-35
-0.5097328
+-7.013437e-07
--3.878958e-34
+-4.0569785e-06
--3.7806562e-34
+-1.6326982e-07
--7.8518477e-35
+-5.024649e-09
--3.8417675e-35
+1.0218174e-08
-5.504886e-34
+-5.870887e-17
--4.2347166e-34
+2.4735778e-05
-3.77638e-34
+-1.8678125e-28
--6.449212e-34
+-5.81377e-18
-3.723454e-34
+4.2194547e-08
--3.4782797e-34
+-8.974654e-09
--7.3213066e-35
+-8.819107e-18
--4.4892873e-34
+-4.0505355e-36
-3.4874208e-34
+1.4275389e-15
-1.1700748e-33
+2.5213077e-35
--1.1355761e-34
+0.0455311
-1.1225075e-33
+-0.16154826
-8.598829e-34
+-0.12516226
--4.3217242e-35
+-0.15351692
--2.7770687e-34
+-0.15327676
--4.541627e-34
+-0.101601385
-2.895937e-34
+-0.09675424
-5.4065008e-34
+-0.009882243
-8.211584e-34
+0.14380045
-2.4092055e-34
+0.17609678
--1.1384675e-33
+0.15136641
-6.7052264e-34
+0.18814708
--8.305206e-34
+0.14553012
--1.8370869e-34
+0.08837449
--5.012333e-34
+-0.033248488
-9.2541105e-34
+-5.5479194e-13
-5.402706e-34
+-1.301003e-35
--3.0262877e-34
+6.407329e-35
-7.088514e-34
+1.9132001e-35
--1.7485143e-34
+9.564731e-15
--2.366834e-34
+-4.683806e-19
-6.501108e-34
+1.8975264e-24
--5.722031e-34
+2.1182613e-16
-1.1429626e-33
+-3.6244807e-35
-4.9021696e-35
+8.7545505e-28
-1.1040688e-34
+6.0832183e-21
--1.0464325e-33
+-8.545767e-31
-1.6525106e-34
+5.2584422e-14
--3.9707304e-34
+2.1925994e-22
-2.1401144e-34
+-3.8261837e-20
--0.051736742
+-1.0880043e-34
--0.042417962
+-6.6652585e-35
-0.051013805
+-6.389439e-35
-0.16345194
+3.2505208e-35
-0.5187456
+-4.973718e-35
--0.14417858
+-3.3143227e-35
--0.0539816
+-8.6873607e-35
--0.15638705
+-7.993331e-35
--0.30926377
+-7.852833e-36
-0.42976364
+-6.3270696e-35
--0.029886993
+5.258114e-35
+2.1151958e-35
+7.3324824e-35
+7.1793427e-35
+-9.0051764e-35
+0.3847243
+-0.008498104
+0.030268772
+-0.13264672
+0.030948505
+-0.07938414
+-0.04668712
+-0.16404602
+0.07313376
+0.1522345
+0.00048681977
+0.01413009
+-0.09327267
+-0.055540953
+-0.100918815
+-0.06997617
+-0.100417346
+-0.11185439
+-0.108010836
+-0.1542093
+-0.026418801
+-0.0976361
+0.2631115
+0.110037416
+-0.038920112
+0.03310242
+-0.07849237
+0.087744445
+-0.016706102
+0.42764086
+-0.063509755
+-0.20127158
+0.3405362
+0.10242782
+-0.25828084
+-0.18461828
+0.18166192
+-0.13771532
+-0.14198124
+0.4270196
+0.16850737
+-0.13088605
+-0.18872121
+0.22758731
+-0.07991603
+-0.06114433
+0.22688313
+0.013428835
+-0.12416983
+0.32349384
+-0.081210054
+-0.33148897
+-0.1045747
+0.20436902
+0.018065251
+-0.15008682
+0.3795789
+-0.022265602
+-0.2928385
+0.012199368