Method, system and computer programs for the automatic counting of the number of insects in a trap
A method and system are proposed for the automatic counting of the number of insects in a trap. The method comprises acquiring, using an acquisition system, a plurality of spectral images of a trap or of a portion of the trap, the spectral images being acquired for at least two different quasi-monochromatic spectral ranges after having sequentially illuminated the trap, or portion of the trap, with light at said two quasi-monochromatic spectral ranges. The trap, or portion of the trap, contains a series of objects adhered thereto, including insects, and optionally other particles. In addition, the method comprises counting, using a processor, the number of insects of a first type through the detection and differentiation of the insects of the first type taking spectral and morphological parameters thereof into account.
1. A method for the automatic counting of the number of insects in a trap, the method comprising:
acquiring, by an acquisition system, a plurality of spectral images of a trap or of a portion of the trap, the plurality of spectral images being acquired for at least two different quasi-monochromatic spectral ranges, a first quasi-monochromatic spectral range and a second quasi-monochromatic spectral range, after having sequentially illuminated the trap, or portion of the trap, with light at said two different quasi-monochromatic spectral ranges, the second quasi-monochromatic spectral range comprising wavelengths greater than the wavelengths of the first spectral range, and the trap, or portion of the trap, containing a series of objects adhered thereto, the objects including insects of one or several types, and optionally other particles including petals, leaves, dust or other types of dirt; and
counting, by a processor, the number of insects of a first type of said insects included in the trap, or portion of the trap, through the detection and differentiation of the insects of said first type taking spectral and morphological parameters thereof into account, said counting comprising:
executing a first algorithm on at least one of the acquired spectral images in one of the two different quasi-monochromatic spectral ranges, wherein the first algorithm comprises applying an intensity threshold to said spectral image, providing a first mask as a result with values equal to 1 for the image pixels corresponding to the background and equal to 0 for the image pixels which comprise objects;
executing a second algorithm on said first mask, wherein the second algorithm comprises applying an area threshold to a number of areas of the first mask with pixel values equal to 0 and assigning a value of 1 to the areas with an area value below said area threshold, providing a second mask as a result;
obtaining a REDIN image by relating on a pixel-by-pixel base intensity values of both acquired spectral images in the quasi-monochromatic spectral range corresponding to short and long wavelengths, respectively;
applying the second mask on the obtained REDIN image, providing a new image; and
executing a third algorithm on said provided new image, wherein the objects of the trap, or portion of the trap, with a value of the REDIN image greater than or equal to a set threshold value are considered insects of said first type.
2. The method according to claim 1 , wherein the third algorithm also comprises applying a pixel connectivity threshold to the provided new image, establishing an area range associated with the insect of the first type, and using said pixel connectivity threshold for considering if an insect is of the first type or not, as long as its size falls within the established area range.
3. The method according to claim 1 , further comprising executing a fourth algorithm, based on eccentricity, wherein the method comprises removing from said new image the objects with an eccentricity value greater than a given eccentricity threshold value.
4. The method according to claim 1 , wherein before the counting of the insects of the first type the method comprises applying a conditioning algorithm on the acquired spectral images, wherein the conditioning algorithm comprises:
calculating a number of reflectances of the first type of insects from a number of intensity values of each pixel of the acquired spectral images taking into account pixel intensities of: an original spectral image of the trap, a dark image of the trap and a spectral image of a reference target; or
carrying out a calibration of said acquisition system, before the acquisition of the plurality of spectral images, wherein the calibration comprises setting, at least, the same acquisition parameters of the acquisition system for said two different quasi-monochromatic spectral ranges.
5. The method according to claim 1 , wherein the first algorithm comprises the Otsu, entropy or k-mean methods.
6. The method according to claim 1 , wherein the first algorithm is carried out using the spectral image corresponding to the first quasi-monochromatic spectral range.
7. The method according to claim 1 , wherein the illumination and the acquisition are carried out in a direction perpendicular to the trap, or portion of the trap, or with a certain angle relative to the trap, or portion of the trap.
8. The method according to claim 1 , wherein the first type of insect is the California red scale.
9. The method according to claim 1 , wherein the first quasi-monochromatic spectral range is comprised between 300-500 nm and the second quasi-monochromatic spectral range is comprised between 600-900 nm.
10. A system for the automatic counting of the number of insects in a trap, comprising:
a trap which contains a series of objects adhered thereto, wherein the objects include insects of one or several types, and optionally other particles including petals, leaves, dust or other types of dirt;
an illumination device configured to sequentially emit light towards the trap, or towards a portion of the trap, in at least two quasi-monochromatic spectral range or at least two illumination devices configured to sequentially emit light towards the trap, or towards a portion of the trap, in at least one quasi-monochromatic spectral range;
an acquisition system, operatively connected to said illumination device(s), and configured to acquire a plurality of spectral images of the trap, or of the portion of the trap, wherein the plurality of spectral images are acquired for at least two different quasi-monochromatic spectral ranges, a first quasi-monochromatic spectral range and a second quasi-monochromatic spectral range, the second quasi-monochromatic spectral range comprising wavelengths greater than the wavelengths of the first spectral range; and
a computing unit including one or more processors and at least one memory, wherein said one or more processors are adapted to count the number of insects of a first type of said insects included in the trap, or portion of the trap, through the detection and differentiation of the insects of said first type taking spectral and morphological parameters thereof into account by means of:
executing a first algorithm on at least one of the acquired spectral images in one of the two different quasi-monochromatic spectral ranges, wherein the first algorithm comprises applying an intensity threshold to said spectral image, providing a first mask as a result with values equal to 1 for the image pixels corresponding to the background and equal to 0 for the image pixels which comprise objects;
executing a second algorithm on said first mask, wherein the second algorithm comprises applying an area threshold to a number of areas of the first mask with pixel values equal to 0 and assigning a value of 1 to the areas with an area value below said area threshold, providing a second mask as a result;
obtaining a REDIN image by relating on a pixel-by-pixel base intensity values of both acquired spectral images in the quasi-monochromatic spectral range corresponding to short and long wavelengths, respectively;
applying the second mask on the obtained REDIN image, providing a new image; and
executing a third algorithm on said provided new image, wherein the objects of the trap, or portion of the trap, with a value of the REDIN image greater than or equal to a set threshold value are considered insects of said first type.
11. The system according to claim 10 , further comprising one or more polarizers arranged in front of at least one of the acquisition system or the illumination device(s).
12. The system according to claim 10 , further comprising a scanning system, operatively connected to at least one of the illumination device(s) or the acquisition system, to carry out a sequential scanning of other portions of the trap.
13. The system according to claim 10 , wherein the first type of insects is the California red scale and wherein the first quasi-monochromatic spectral range is comprised between 300-500 nm and the second quasi-monochromatic spectral range is comprised between 600-900 nm.
14. A non-transitory computer readable medium including code instructions, which, when implemented in a computing device, execute a method according to claim 1 .