def run(self):
print(" Start capture ...")
dmcam.cap_start(dev)
# blocking code goes here
self.run = True
while self.run:
global f_dist, f_gray, frame_data, f_mutex
if dev is None:
time.sleep(0.5)
# get one frame
finfo = dmcam.frame_t()
ret = dmcam.cap_get_frames(dev, 1, frame_data, finfo) # will blocking wait
# ret = dmcam.cap_get_frame(dev, frame_data, None)
# print("get %d frames" % ret)
if ret > 0:
w = finfo.frame_info.width
h = finfo.frame_info.height
# print(" frame @ %d, %dx%d (%d)" %
# (finfo.frame_info.frame_idx,
# finfo.frame_info.width,
# finfo.frame_info.height,
# finfo.frame_info.frame_size))
f_mutex.lock()
dist_cnt, f_dist = dmcam.frame_get_distance(dev, w * h, frame_data, finfo.frame_info)
gray_cnt, f_gray = dmcam.frame_get_gray(dev, w * h, frame_data, finfo.frame_info)
if dist_cnt != w * h:
f_dist = None
if gray_cnt != w * h:
f_gray = None
f_mutex.unlock()
self.frameReadySignal.emit(0)
self.captureDoneSignal.emit(int(self.run))
def poll_frame(self):
finfo = dmcam.frame_t()
ret = dmcam.cap_get_frame(self.dev, self.frame_data, finfo)
if ret > 0:
w = finfo.frame_info.width
h = finfo.frame_info.height
_, frame_dist = dmcam.frame_get_distance(self.dev, w * h, self.frame_data, finfo.frame_info)
_, frame_gray = dmcam.frame_get_gray(self.dev, w * h, self.frame_data, finfo.frame_info)
self.frame_dist.append(frame_dist.reshape(h, w))
self.frame_gray.append(frame_gray.reshape(h, w))
if len(self.frame_dist) > 4: self.frame_dist.pop(0)
if len(self.frame_gray) > 4: self.frame_gray.pop(0)
self.frame_cnt += 1
return ret > 0
def tof_main(head_alarm):
"""
ToF 主函数
"""
finfo = dmcam.frame_t()
ret = dmcam.cap_get_frames(dev, 1, f, finfo)
# print("get %d frames" % ret)
if ret > 0:
w = finfo.frame_info.width
h = finfo.frame_info.height
dist_cnt, dist = dmcam.frame_get_distance(dev, w * h, f,
finfo.frame_info)
gray_cnt, gray = dmcam.frame_get_gray(dev, w * h, f, finfo.frame_info)
if dist_cnt == w * h:
# -----------------------------------------
dep0 = (dist.reshape(h, w) * 1000)
Z = gray.reshape(h, w)
img_dep = cv2.convertScaleAbs(dep0, None, 1 / 16)
img_to_pred = img_dep[:, cut_edge:320 - cut_edge]
img_to_pred = cv2.merge([img_to_pred] * 3)
img_show = img_to_pred.copy()
# cv2.imwrite('ToF'+'/dep_{}.png'.format(time.time()),img_show)
predict_label = 0
if use_CNN_model:
img_to_pred = np.expand_dims(img_to_pred, axis=0)
y_pred = model.predict(img_to_pred)[0]
if head_alarm == 1:
y_pred[1] += 0.2
predict_label = np.argmax(y_pred)
max_prob = np.max(y_pred)
if max_prob < 0.55:
predict_label = 0
show_label = dl.after_delay(predict_label)
return img_show, show_label
}
wparams[dmcam.PARAM_ILLUM_POWER].illum_power.percent = pwr_percent
wparams[dmcam.PARAM_INTG_TIME].intg.intg_us = 1000
wparams[
dmcam.PARAM_FRAME_FORMAT].frame_format.format = dmcam.DM_FRAME_FMT_DISTANCE
if not dmcam.param_batch_set(dev, wparams):
print(" set parameter failed")
print(" Start capture ...")
dmcam.cap_start(dev)
f = bytearray(320 * 240 * 4 * 2)
run = True
while run:
# get one frame
finfo = dmcam.frame_t()
ret = dmcam.cap_get_frames(dev, 1, f, finfo)
# print("get %d frames" % ret)
if ret > 0:
print(" frame @ %d, %dx%d (%d)" %
(finfo.frame_info.frame_idx, finfo.frame_info.width,
finfo.frame_info.height, finfo.frame_info.frame_size))
# print the first 16bytes of the frame
# print([hex(n) for n in f][:16])
w = finfo.frame_info.width
h = finfo.frame_info.height
gray_cnt = 0
dist_cnt, dist = dmcam.frame_get_distance(dev, w * h, f,
finfo.frame_info)
def get_frames(self):
if self.dev is None:
log.warning("Device not initialized, retry to capture in 0.5s!")
time.sleep(0.5)
return
f = bytearray(320 * 240 * 4 * 2)
finfo = dmcam.frame_t()
log.debug("dmcam.cap_get_frames")
ret = dmcam.cap_get_frames(self.dev, 1, f, finfo)
w = finfo.frame_info.width
h = finfo.frame_info.height
intg_auto_arg = dmcam.filter_args_u()
intg_auto_arg.sat_ratio = 5
if not dmcam.filter_enable(
self.dev, dmcam.DMCAM_FILTER_ID_AUTO_INTG, intg_auto_arg,
sys.getsizeof(intg_auto_arg)):
print(" enable AUTO INTG filter (sat_ratio=%d) failed" %
intg_auto_arg.sat_ratio)
if ret <= 0:
if ret == -10:
log.warning("dmcam USB cable unplugged, try to reconnect!")
self._close_dmcam()
self._init_dmcam()
self._start_dmcam()
elif ret == 0:
self.empty_frame_cnt += 1
if self.empty_frame_cnt > 20:
log.warning("dmcam capture timeout, restart capture")
self._stop_dmcam()
self._start_dmcam()
self.empty_frame_cnt = 0
else:
log.warning("dmcam capture error, errono:{}".format(ret))
self._stop_dmcam()
self._start_dmcam()
else:
self.empty_frame_cnt = 0
print()
dist_cnt, dist = dmcam.frame_get_distance(self.dev, w * h, f,
finfo.frame_info)
# gray_cnt, gray = dmcam.frame_get_gray(self.dev, w * h, f, finfo.frame_info)
# gray_data = gray.astype(np.uint16)
# gray = ((gray/16).astype(np.uint8)).reshape(240,320)
if dist_cnt == w * h:
data = (dist * 1000).astype(np.uint16)
frame = data.reshape(240, 320)
cv2.imshow('frame', cv2.convertScaleAbs(frame, None, 1 / 20))
cv2.waitKey(1)
# ## 静态手势
img_s = frame.copy()
result_s, result_str = static(img_s)
self.result_list = update_win(self.result_list,
result_s,
win_len=5)
## 动态手势
img_d = frame.copy()
img_d = img_d.astype(np.float32) / 10000
self.algo = mainloop_trk(self.algo, img_d, self.viewer,
self.bg)
self.viewer.put_text(str('手势识别结果为：'), 150, 1000)
if result_s > 0 and self.algo.decision == 'point':
if self.ENABLE_VIEW:
# if len(self.result_list)>0:
# counts = np.bincount(self.result_list)
# idx = np.argmax(counts)
# if idx > 0:
# self.viewer.put_text(str(label[idx]),200,1050)
self.viewer.put_text(str(result_str), 200, 1000)
# final_result = result_str
else:
x, y = self.algo.xlist[self.algo.idx], self.algo.ylist[
self.algo.idx]
if self.ENABLE_VIEW:
if self.algo.trace_pattern['type'] != 'point':
self.viewer.draw_circle(int(x * 2 + 400),
int(y * 2 + 165),
r=5,
line_wid=5)
draw_trace_mode(
self.viewer,
mode=self.algo.trace_pattern['type'],
pos=self.algo.trace_pattern['param'])
if self.algo.trace_pattern['type'] == 'circle':
self.viewer.put_text(str('转圈'), 215, 1050)
elif self.algo.trace_pattern['type'] == 'h-line':
self.viewer.put_text(str('水平移动'), 215, 1050)
elif self.algo.trace_pattern['type'] == 'v-line':
self.viewer.put_text(str('竖直移动'), 215, 1050)
# self.result_list=[]
# if self.algo.trace_pattern['type'] == 'point':
# final_result = 'None'
# else:
# final_result = self.algo.trace_pattern['type']
# if final_result == "circle":
# pos = self.algo.trace_pattern['param']
# update_win(pos_win, pos, win_len=3)
# if pos_win[0] < pos_win[1] and pos_win[1]<pos_win[2]:
# final_result = "c-circle"
# elif pos_win[0] > pos_win[1] and pos_win[1] > pos_win[2]:
# final_result = "circle"
# if final_result == "h-line":
# pos = self.algo.trace_pattern['param']
# update_win(pos_win, pos, win_len=3)
# if pos_win[0] < pos_win[1] and pos_win[1]<pos_win[2]:
# final_result = "right"
# elif pos_win[0] > pos_win[1] and pos_win[1] > pos_win[2]:
# final_result = "left"
# if final_result == "v-line":
# pos = self.algo.trace_pattern['param']
# update_win(pos_win, pos, win_len=3)
# if pos_win[0] < pos_win[1] and pos_win[1]<pos_win[2]:
# final_result = "down"
# elif pos_win[0] > pos_win[1] and pos_win[1] > pos_win[2]:
# final_result = "up"
# 刷新屏幕显示
if self.ENABLE_VIEW:
self.viewer.update()
evt, param = self.viewer.poll_evt()
if evt == 'quit':
return False
return True
