def predict(self, bbox, time, fid):
bbox = bbox[:4]
if fid == 0:
self.pre_bbox = np.array(bbox)
bbox = Exponentially_weighted_averages(self.pre_bbox,
bbox,
fid,
theta=0.8)
x, y = bird_view_proj((bbox[0] + bbox[2]) / 2, bbox[3], self.K,
self.dist_coeff, self.R, self.T)
x = np.sqrt(x * x - 3.8**2)
interval = 5
if fid == 0:
self.pre_vx = 0.
self.pre_bbox = np.array(bbox)
time_diff = 0x3ffffff
elif fid >= interval:
time_diff = time - self.times[-interval]
self.pre_x = self.result[-interval]['x']
else:
time_diff = time - self.times[0]
self.pre_x = self.result[0]['x']
self.times.append(time)
vx = (x - self.pre_x) / time_diff
if len(self.result) == 1:
self.pre_vx = vx
vx = Exponentially_weighted_averages(self.pre_vx, vx, fid, theta=0.9)
self.pre_bbox = bbox
self.pre_x = x
self.pre_vx = vx
cur_pred = {
'fid': fid,
'vx': vx,
'x': x,
# 'ref_bbox': {
# 'left': float(bbox[0]), 'top': float(bbox[1]),
# 'right': float(bbox[2]), 'bot': float(bbox[3])
# }
}
self.result.append(cur_pred)
return cur_pred
def proj(self, u, v):
x, y = bird_view_proj(u, v, self.K, self.dist_coeff, self.R, self.T)
return x, y